Mistral releases Mistral Large 4 ("Le Chonk") weights: leading open-source model in coding, still behind Chinese models overall
Paris-based Mistral releases "Le Chonk," a strong coding model that still trails China overall
AI summary
Mistral released the weights of its largest model, Mistral Large 4 ("Le Chonk"), with about 1 trillion parameters and 52B activated per step. Terminal-Bench 4.0 rose from 28% to 41%, slightly above GLM-5.3's 40%, and FrontierCode 2.0 ranked first among open models at 19.7%, but Claude Opus 5.5 scored 46%. It scored only 26 points on the Artificial Analysis Intelligence Index v5, behind DeepSeek V4.1 Flash and GLM-5.3 at 32 points; API pricing is $0.70 per million input tokens and $2.10 per million output tokens.
Read original